I would like to know if you are considering the “doming effect” for DEM-generation? It appears, when taking images from near-parallel directions and causes systematic errors in the 3D-model (DEM). See this paper for more information:
Therefore, some convergent images may help avoid this problem. Couldn’t this be done, using the center of the area to circle around and take some additional images?
Hi,
Thanks for the informative post - a really interesting area we are actively working on! We are definitely considering the “doming effect” in our DEMs - as you mentioned, near-parallel flight plans do ofter exhibit this effect, and it is especially pronounced for narrow mapping regions (where maybe only 2 or 3 parallel paths are flown), e.g. road surveys.
As the article suggests, we do often see improved results when oblique images are included along with the nadir set, so we would recommend this approach for accurate 3D modeling. The oblique images also really help with accurate reconstruction of any 3D objects in your mapping area (e.g. towers, buildings, etc.); we have some documentation on capturing oblique imagery to supplement your automated DroneDeploy flight here which is hopefully useful to you.
We are actively working on improving the data post-processing to reduce these bowling / doming effects, and will keep you updated with our progress.
Thanks again for your post and the linked article - a great read!
Jeremy
My work includes surveying ancient linear fortifications from Roman ages called vallums. I flew a mission over one such monument and the DSM reported in the middle values like -53 m on a very flat terrain. The mission’s DSM is here
https://www.dronedeploy.com/app/data?planId=566d3bb35c70cb000a6bd93a
So there is a possible way to counteract that by flying nadir missions… Wish I knew that last weekend… The objective was some 500 km from where I live so going back to it takes half a day and a full tank of diesel. It would be really nice to have DD automatically tilt the camera for us at 45 degrees now and then while doing survey missions… as it will vastly improve the 3d model creations as well. There are apps like Litchi who are able to perform such feats…
Hi, I recently mapped a farm with both the free trial of DD and Pix4D. I ws was quite surprised to see the difference in DSM outputs from the two software.
Here is the DSM output from PIX4D with the same photoset. Both taken with a Inspire 1 PRO.
Google “Gently Curved, Convergent, Non-traditional Drone Flight Paths”.
Curved flight paths may help mitigate the SfM doming (elevation) error - especially for oblique imagery!
Hi @James_Dow
I read your article here.
Is this your work (patent request, ebook) ?
Have you any test projects to share comparing the results from an image set collected with the typical lawn-mower pattern vs. a curved track?
It was actually published in 2014. This one goes into a little more detail, but I think @James_Dow article references better on how to overcome it in a simpler manner.
Dave: Yes, this is my work. The work evolved over the past couple of years attempting to prove that the SfM collection and processing of imagery data could, in fact, replace LIDAR for solutions of the electric power transmission vegetation management program. During the development of the curved flight line method, more than 100 miles of electric power transmission rights-of-way were repeatedly flown and mapped. The imagery was processed using AgiSoft, Pix4D, OneButton, and Zephyr; and improvements in the elevation values improved significantly - using both nadir and oblique imagery. The x/y values were not so much an issue.
At this point, I am not interested in sharing the huge volume of data, as it is not organized for sharing or any published work at the present.
The flight paths of one of the electric power transmission corridors is illustrated in the ebook.
Yes, a patent application has been filed for this work.
Jim Dow
Dave: I forgot to mention that about 5 miles of highway corridor and about 2 square miles of urban environment were also evaluated - repeatedly.
Jim Dow
Thanks for that, Jim.
Using dji craft, drawing up curved line missions would need to be done in Litchi and maybe AutoPilot at this point which would take some time. Alternatively, using DD or Map Pilot, the camera could be set slightly oblique at 80* rather than 90* (nadir) or one mission with nadir and another at an angle or cross hatch with a slightly oblique gimbal setting.
What are your thoughts at this point?
Has anyone ever see this with crosshatch? Even at 90?
All you have to do in Litchi is the normal lawnmower, put a waypoint in the middle of each leg and set your radius to 90%.
I should have qualified by saying compared to DD or MP, setting up a mission in Litchi is more time consuming. In addition, you would need to set the camera to take pictures based upon time rather than distance traveled. Not a huge difference, but can be meaningful from what I understand.
If using curved legs is truly magnitudes better than straight legs, it seems like it would be fairly easy for the devs to have the app insert a mid-point and offset it with a curve value just like Litchi does.
Agreed simpler, but in progressive mission you set it once and fly. Two-second time interval always which is a standard camera setting. Adjust your altitude and speed
Dave: Here is some research (other than my own) that substantiates the effectiveness of using curved flight lines along with convergent oblique photos. Please note that the use of curved flight lines is particularly effective for longer, narrower target areas. Just to illustrate the use of curved flight lines for larger areas, I’ve included two illustrations: a) traditional (linear/parallel) flight lines over a glacier and b) curved flight lines over the same glacier.
Minimising systematic error surfaces in digital elevation models using oblique convergent imagery
First published: 16 March 2011
https://doi.org/10.1111/j.1477-9730.2011.00623.x
Results of the simulation process, the laboratory test and the practical test are reported in this paper and demonstrate that an oblique convergent image configuration eradicates the systematic error surfaces which result from inaccurate lens distortion parameters. This approach is significant because by removing the need for an accurate lens model it effectively improves the accuracies of digital surface representations derived using consumer‐grade digital cameras. Carefully selected image configurations could therefore provide new opportunities for improving the quality of photogrammetrically acquired data.
Note: The curved flight lines can be combined (with the eccentricity altered to accommodate the particular aircraft/drone used and/or pilot preference) to create particularly efficient flight paths - with no turnout loops. Refer to the third example/illustration.
This last example is for an open-pit copper mine with an area of about one square mile. It can be flown a) end-to-end by combining opposing convergent flight lines or b) using complete loops - flying complete circuits.
Dave: Here’s an example of adapting curved flight lines to capture images along long, narrow corridors (electric power transmission corridors). The fixed-wing drone was flying at 31mph with a height of 400’AGL. The target corridor is about one mile in length and a width of 2200’. No flight line segment is linear; and no flight line segment is parallel to any other segment. Therefore, no to images have the same orientation. The objective is to mitigate the systematic SfM doming (elevation) error.
Just to keep this topic going, I was approached by a drone (multi-rotor) operator in Quebec, Canada asking me to prepare a curved flight line mission plan for a 1.5 sq mile mine site in northern Quebec. I responded with the attached mission plan - note there are no turnout loops required for a multi-rotor. The operator plans to fly this area using both the curved flight lines and the traditional linear/parallel flight lines and compare the results. So, in the near future, we should have an independent assessment.
James, not a DJI bird I suspect? Isn’t there a max. # of waypoints issue if trying to incorporate curved missions of any size from the DJI sdk? At least that is what the Map Pilot devs stated when asked about this. They suggested sufficient overlap is all that is required and if one were concerned about it then a double grid at less than 90* (not cross hatch) would also kill the doming effect. It will be interesting to see what the results from this test present. FWIW, most big commercial drone surveying outfits are just suggesting single grid missions with 75/75 overlap and GCP and checkpoints. No cross-hatch, obliques, or non-linear capture methodology.
I would offer you all another example of the use of curved drone flight lines. Here’s a proposed mission plan for collecting drone (fixed wing) imagery for an airport obstruction assessment - illustration.
James, do you have a practical flight programming solution, particularly using DJI craft?
